ZF-1276: add ignore_user_abort to Zend_Cache_Backend_File
Description
If the user disconnect during the write process the cache contains not clompleted data. This can shirk by set the ignore_user_abort to off and after writing to prior value again.
If the user disconnect during the write process the cache contains not clompleted data. This can shirk by set the ignore_user_abort to off and after writing to prior value again.
Comments
Posted by Gavin (gavin) on 2007-04-16T13:04:20.000+0000
I suggest considering adding this as an option to Zend_Cache.
Posted by Fabien MARTY (fab) on 2007-04-16T15:33:18.000+0000
IMHO, this is useless.
With the default configuration, the hash key won't be ok so the cache will be dropped.
So it's not really a problem if the cache contains not completed data because it won't be used (readControl feature)
Posted by Gavin (gavin) on 2007-04-16T16:03:52.000+0000
True, but in some situations, a failed cache write could be avoided using ingore_user_abort().
Posted by Fabien MARTY (fab) on 2007-04-16T16:33:01.000+0000
sure but is it enough to add a new option in Zend_Cache core ?
I prefer to keep the core away from "not really usefull" options because : - the code is cleaner - perfs are better
Posted by Gavin (gavin) on 2007-04-16T16:47:33.000+0000
I don't have any strong feelings on this, and won't object if you close the issue, because developers could call ignore_user_abort() themselves. However, I doubt adding the following would affect perfs in any meaningful way for those not enabling the proposed option:
On the other hand, if something slow and costly was calculated and written by Zend_Cache* corrupted because of a user abort not ignored, the perf cost could be large.
Posted by Fabien MARTY (fab) on 2007-04-19T14:52:17.000+0000
I don't know. I prefer to take some time to think about it.
Posted by Thomas Weidner (thomas) on 2007-05-06T09:30:25.000+0000
Zend_Cache related issue, assigned to Fabien until he has thought about it ;-)
Posted by Fabien MARTY (fab) on 2007-05-08T07:49:34.000+0000
ok for 1.1.0
Posted by Fabien MARTY (fab) on 2008-01-13T12:48:10.000+0000
fixed in SVN